I measured the inductance of the compressor - 43mH. I also measured the current drawn by the motor - 0.66 Amps. Then I calculated the energy stored in the inductance, which is "a half L I squared", or 0.5*L *I^2. There is a similar equation, "a half C V squared", (0.5C*V^2) for the energy stored in a capacitor, and I reasoned that I needed a capacitor to store about the same energy as was in the inductor. It worked out at 315nF. I had a couple of 220nF capacitors in the junk box. Importantly they were X2 rated capacitors (see the marking in the photo below) which are suitable for connecting directly across the mains.
When I looked on the internet there was a lot of stuff about designing snubbers for MOSFET circuits, and Switchmode powersupply circuits. They all said that the time constant of the C and the R needed to be about 10% of the "on time". Not sure how that relates to a thermostat, but the mains frequency is 50Hz, and so each cycle is 0.02 Sec. The time constant of my 220nF cap with a 10k resistor is about 0.002 Sec so I plumped for a 10k resistor. The time constant is calculated by multiplying the C by the R. And it is the time taken for the capacitor to discharge to about 30% of it's starting Voltage when discharging through the resistor. It probably doesn't matter much as long as there is some resistance to dissipate the energy.
When the motor is not running the snubber will have quite a lot of Voltage across it - most of the mains Voltage. A 220nF capacitor will pass quite a few milliamps at 240V. It is possible to work it out with phasor diagrams and power-factors and things, but I simply connected the circuit across the mains and measured the Voltage across the resistor - which was about 135V. Since power = V squared over R, that is about 1.8 Watts being dissipated in the resistor. In fact it will be less, because the motor is in series. At first I paralleled up 10 x 100k metal film resistors, but that is quite a fat bundle. I found it was better to series 10 x 1k resistors and cover them with heatshrinkable tubing. Each resistor is good for about 0.6W, so in theory my chain should be ok dissipating 6W, but it is good to have some margin of safety. The long thin arrangement dissipates the heat better. I put five on each leg of the capacitor.
So I had an arrangement like this:
The Nordfrost thermostat is mounted in a plastic box in the front of the cabinet, so it is quite easy to get at - there is also a fair amount of room inside the box to accommodate extra components. I used some of the "piggy-back" type 1/4" push-on connectors to connect the snubber circuit. Since these just crimp on to the wires there was no soldering to do. Here's a picture with the cover removed - there is rather a tangle of wires, but nothing is danger of shorting to anything else so it is ok!
The other slight downside to the snubber circuit is an increase in power consumption - the result of having a small current flowing in the resistor when the motor is off - I guess it is about a Watt of electricity wasted during the off period of the freezer.
Perhaps an in-line RF filter might help with the TV interference.
At least the snubber should extend the life of the thermostat contacts.